It's known that with the finger release of the recurve archer the nock end
of the arrow is accelerated sideways, away from the bow, and this sideways
acceleration initiates the Archer's Paradox effect. As far as I know no research
has been done on how this sideways acceleration is generated. The following
are two suggested mechanisms for generating this sideways acceleration that
I'll call the slingshot effect and the tab slide effect.
Slingshot Effect
With the slingshot effect the string is located in the second nuckle joint
and a 'deep hook' finger configuration is used when holding the arrow at full
draw.
The key point is that the string, with respect to the direction of
the shot, is located to the rear of the first knuckle joint. At the loose
the string stays in the same place on the tab (in the second knuckle joint)
and the fingers are rotated by the string, the axis of rotation being more
or less the first knuckle joint. The string therefore travels in the arc
of a circle forwards and sideways. These resulting forward and sideways displacements
appear to be reasonably compatible with those required to obtain the typical
Archer's Paradox behaviour. The diagram illustrates the process.
Tab Slide Effect
In this case the fingers have much less of a hook and the string is located
in front of the first nuckle joint.
The string is mainly restrained against
the pads of the flesh of the fingertips. At the release the fingers are again
rotated more or less around the first nuckle joint and the string slides
over the surface of the tab. It is the force interaction between the tab
and the string that provides the sideways nock acceleration. The following
diagram illustrates the process.
Kooi's arrow model suggests that the displacement of the nock by the fingers is around 2mm sideways movement over 3mm forwards travel. Both release mechanisms appear to be compatible with this required nock movement. Depending on how much hook there is in the fingers (angle a) it is feasible
that the loose is a combination of the two processes. It should be noted
that in both cases the string has left the tab long before it gets anywhere
near the fingertips.
Most tabs will develop some permanent curl in them so although the fingers
may be out of the way the string can be dragging across or hitting the outer
edge of the tab material on the way past.
The key point is that the string, with respect to the direction of
the shot, is located to the rear of the first knuckle joint. At the loose
the string stays in the same place on the tab (in the second knuckle joint)
and the fingers are rotated by the string, the axis of rotation being more
or less the first knuckle joint. The string therefore travels in the arc
of a circle forwards and sideways. These resulting forward and sideways displacements
appear to be reasonably compatible with those required to obtain the typical
Archer's Paradox behaviour. The diagram illustrates the process. 
The string is mainly restrained against
the pads of the flesh of the fingertips. At the release the fingers are again
rotated more or less around the first nuckle joint and the string slides
over the surface of the tab. It is the force interaction between the tab
and the string that provides the sideways nock acceleration. The following
diagram illustrates the process.